Abstract: With the widespread access of renewable energy sources such as photovoltaic, wind, and solar to the grid, there is an urgent need for electric utilities to implement accurate short-term load forecasting to ensure the stable operation of the grid. In this paper, a data decomposition technique is employed to eliminate noise and random disturbances in the load data, and a variational modal decomposition(VMD) algorithm is introduced to decompose the original load sequence into simple subsequences of different frequencies. Based on these subsequences, a combined prediction method combining VMD and improved CNN-LSTM is proposed in this paper. From the example analysis, it is shown that the RMSE, MAPE, and MAE indexes of the VMD-DA-RCLSTM model are reduced, which indicates that the proposed combined forecasting model helps to improve the accuracy of power load forecasting.